JeecgBoot versions through 3.9.5 have a missing authorization vulnerability in the AiragAppController release handler. This flaw allows any authenticated user, including low-privileged attackers, to publish or unpublish AI applications belonging to other users. Attackers can send POST requests to the /airag/app/release endpoint to obtain share tokens, which expose applications to anonymous chat access, or to invalidate existing share links. The vulnerability has a medium severity rating with a CVSS score of 5.4.
### Summary
The Pydantic AI development web chat UI (`Agent.to_web()`, `clai web`) did not check the content type of requests to its chat endpoint. This allowed a website visited by a developer to submit a request to a chat UI running on that developer's machine, causing the served agent to run and to execute its tools with the privileges and credentials of the local process.
Binding the web UI to localhost — the default — does not prevent this, because a page open in the developer's browser can reach the loopback address.
### Impact
Applications and developers serving an agent through `Agent.to_web()` or `clai web`. The consequences depend on the tools the served agent exposes, and can include data disclosure as well as unwanted tool side effects. Tools marked `requires_approval=True` were not protected either, because the endpoint trusts approval decisions relayed by the client.
### Mitigation
Upgrade to a patched version. The chat endpoint now requires `Content-Type: application/json` and rejects other requests before the request body is parsed and before the agent runs. The bundled chat UI already sends this header; scripts and other non-browser clients that call the endpoint directly may need to be updated to send it.
If you cannot upgrade, don't run the web UI while browsing untrusted sites, stop it when you aren't using it, and don't serve an agent with side-effecting tools through it.
### Credits
- Thai Son Dinh from VinSOC Labs (R&D)
This report summarizes multiple cybersecurity developments including a set of iCloud email spoofing vulnerabilities, phishing campaigns targeting AI policy experts, a Chrome adblocker extension spying on AI chat conversations, and a large batch of vulnerabilities patched by Kiteworks. The iCloud spoofing bugs allowed attackers to send emails from arbitrary icloud.com addresses that passed SPF, DKIM, and DMARC checks due to inconsistent mail header parsing. The phishing campaigns involved a China-aligned group using fake AI advisory committee invitations to steal credentials via an adversary-in-the-middle proxy. The Poper Blocker Chrome extension was found to collect users' browsing history and AI chat conversations stealthily. Kiteworks patched over 100 vulnerabilities, including critical bugs that could lead to account takeover and code execution. Additional highlights include AI-assisted discovery of Android app flaws and a Cloudflare Containers data leakage vulnerability that was patched. No active exploits in the wild were reported for these issues.
n8n versions before 2.39.6 and 2.40.0 before 2.40.1 contain an approval bypass vulnerability in the Send and Wait node's Approve Within Chat mode. Attackers can submit resume requests without verification of the requester's identity or approval permissions, allowing unauthenticated users to advance waiting executions and trigger guarded actions.
CVE-2024-31026 is a critical remote code execution vulnerability in Greek Universities Network (GUnet) Open eClass Platform version 3.15. It allows an unauthenticated attacker to execute arbitrary code by sending crafted input through the chat field in the course module. The vulnerability has a high CVSS score of 9.8, indicating severe impact on confidentiality, integrity, and availability. No patch or official remediation information is currently provided.
vLLM before 0.29.0 accepts user-controlled stop_token_ids on the OpenAI-compatible POST /v1/completions and POST /v1/chat/completions endpoints but validates only that the values are integers, not that each token id is within the model vocabulary/logits range. When min_tokens > 0, the stop token ids are used as logits indices to suppress stop tokens, so an out-of-range id reaches a CUDA indexing operation (index_put_) and triggers a device-side assertion. An authenticated API user can send a single malformed completion request that returns 500 Internal Server Error and puts EngineCore into a fatal state, causing subsequent requests to fail until the service is restarted (denial of service).
OpenClaw (npm package `openclaw`) before 2026.7.1 fails to restrict owner-only infrastructure tools exposed through the chat.send endpoint. In Gateway deployments using authentication modes that honor caller identity and narrower operator scopes, a write-scoped non-owner caller can start a chat turn whose tool inventory includes the `gateway` and `cron` tools, causing the agent to invoke owner-only configuration or scheduling operations, including persistent state changes. Practical impact depends on the tools selected by the model and the caller's ability to steer the turn. Shared-secret token and password callers are treated as fully trusted operators under OpenClaw's security model and are outside the scope of this issue. The issue is fixed in 2026.7.1; as a workaround, restrict chat.send to administrators in identity-bearing deployments and remove `gateway` and `cron` from affected agent tool policies.
OpenClaw versions >= 2026.3.25 and < 2026.8.1 apply invalid-token rate limiting for Synology Chat webhooks before authentication and key the limit on the raw proxy socket address. In deployments where OpenClaw sits behind a trusted reverse proxy or tunnel and multiple external clients share a single socket address, an unauthenticated sender can exhaust the shared invalid-token budget, causing subsequent legitimate Synology Chat webhook callbacks to be rejected until the rate-limit window expires. The attacker cannot obtain a valid token or read message data; the impact is temporary loss of channel availability. Fixed in 2026.8.1.
Novu provides an API for sending notifications through multiple channels. Prior to 3.18.0, Novu accepts chat webhook URLs from subscriber credentials.webhookUrl, channel endpoint endpoint.url, event payload.webhookUrl, and event overrides.webhookUrl, then passes the selected endpoint.url through send-message-chat.usecase.ts to raw HTTP requests in the Slack, Discord, Mattermost, Microsoft Teams, Grafana On-Call, Ryver, Rocket.Chat, GetStream, and Zulip providers. An authenticated user can supply an internal or otherwise restricted destination because these paths do not apply normalizeOutboundHttpUrl, assertSafeOutboundUrl, or the DNS-pinned safeOutboundJsonRequest protection used by the generic webhook providers. The Novu worker can consequently issue attacker-directed POST requests to internal network services and cause interactions or actions supported by those services. This issue is fixed in version 3.18.0.
Cross-Site Scripting (XSS) vulnerability due to inadequate input sanitisation in the client-side rendering engine of the 1millionbot AI Chat Platform. An unauthenticated remote user could cause external hyperlinks to be rendered in the web interface by sending messages containing Markdown syntax and certain unsanitised content blocks. The impact is limited to the user’s own interactive session; no compromise of internal infrastructure, access to third-party data or impact on administrative panels has been identified.
CVE-2026-90564 is a cross-site scripting (XSS) vulnerability in quequnlong shiyi-blog versions 1.0.0 through 1.2.1. The flaw exists in the SysChatMsgMapper.getChatMsgList function within the chat sendMsg endpoint, allowing remote attackers to manipulate the chat_msg argument to execute XSS attacks. The project has been informed but has not yet responded or issued a fix.
n8n is an open source workflow automation platform. Prior to 2.37.7 and 2.38.2, the /chat WebSocket route accepted a resumeToken and resumed a paused execution without checking that the target node supported chat messages. An anonymous form submitter who received that token could reuse it on the chat route to release a Send-and-Wait, non-chat HITL, or Wait approval gate. The affected authorization logic is packages/cli/src/chat/chat-execution-manager.ts, where canResumeOverChat did not gate the resume target. This issue is fixed in versions 2.37.7 and 2.38.2.
n8n is an open source workflow automation platform. Prior to 1.123.76, 2.37.7, and 2.38.2, the OpenAI Chat Model node enforced credential allowed-domain restrictions for normal calls but not for the model-search dropdown. A workflow editor could set options.baseURL to an arbitrary host and make the searchModels path send the openAiApi credential there. The affected implementation is packages/@n8n/nodes-langchain/nodes/llms/LMChatOpenAi/methods/loadModels.ts, which omitted assertOpenAiCredentialAllowsUrl. This issue is fixed in versions 1.123.76, 2.37.7 and 2.38.2.
CVE-2026-19042 is a command injection vulnerability in TeamViewer Full Client and Host for Linux versions 13.0, 14.0, and 15.0. It allows a remote attacker to execute arbitrary commands as the current user by sending a specially crafted URL through the out-of-session chat feature. Exploitation requires user interaction by clicking the malicious link. The vulnerability has a high severity with a CVSS score of 8.8.
A zero-click remote code execution (RCE) vulnerability in the /Terminal/Notification.hs component of SimpleX Chat before v6.5 allows attackers to execute arbitrary commands in the context of the application without user interaction via sending a crafted payload in a text message.
CVE-2026-75575 is a medium severity vulnerability in Rocket.Chat where the sendForgotPasswordEmail method lacks rate limiting. This allows unauthenticated attackers to trigger unlimited password reset emails to any account, potentially enabling email flooding and user enumeration. The vulnerability affects multiple versions prior to specific fixed releases. Later versions implement a rate limit of ten calls per minute per client IP to mitigate the issue.
CVE-2026-71366 is a server-side request forgery (SSRF) vulnerability in multiple AWX notification backends of Red Hat Ansible Automation Platform 2.5 for RHEL 8. Notification backends including webhook, Mattermost, Rocket.Chat, and Grafana use notification template URLs as direct HTTP request targets without validating against private, loopback, or reserved IP ranges. This allows an organization notification administrator to cause the AWX control node to make HTTP requests to internal or loopback addresses. The webhook backend also follows HTTP redirects and resends Basic Authentication credentials to redirected hosts without host validation, risking credential exfiltration. The Grafana backend sends its API key in the Authorization header to the configured URL. No complete mitigation exists except applying the forthcoming update; interim mitigations include restricting notification template creation to trusted admins and network egress filtering.